Design And Implementation Of Face Information Analysis System

Today with the spread of more and more human-computer interactive devices, smart signage, smart phones, tablet PCs, IPAD, have become a part of people’s daily life. Although these devices are equipped with cameras, most of the devices use touch or voice during the interaction instead, face information, such as gender, expression,age, and other information is not well used. So to make the interaction between devices and human become more smart, strong technologies, computer vision, machine learning and artificial intelligence, need to be applied in. In this thesis, we mainly introduce these technologies and integrate them into the development of this face information analysis system.In the paper,we first introduce the compressed sensing theory, which is popular in signal processing filed, and dicsuss the signal sparse representation, measurement matrix and signal reconstruction problems in detail. And we propose a sparse measure matrix that meet some special conditions to extract image feature rapidly.The theory is an important foundation of this system, then to track face efficiently, we use a simple bayesian classifier,which can meet the requirement of a real-time running system. Next we also discuss a machine learning method, support vector machine theory in detail, support vector machine’s excellent classification ability guarantees the system’s gender, expression, age recognition capability, and for the multi-class classification problems, such expression and age of classification, we also explore some different classification strategies.Finally, we integrate face tracking module, gender recognition module, expression recognition module and age recognition module into the system, and the experiment shows that the compressed sensing theory, support vector machines and other methods are effective.